CMDashCybersecurity Maturity Dashboard

← All landscape areas

Emerging / Frontier · Situational

AI / ML Security (AI-SPM)

Secures AI models, pipelines, and LLM apps.

11 options listedTypical cost: $25k-$150k/yrGOVERN (GV) / IDENTIFY (ID) / PROTECT (PR)

Overview

This emerging category secures the AI lifecycle: model/dataset integrity, AI posture management, and runtime guardrails against prompt injection and data leakage for LLM apps. It aligns with NIST AI RMF and ISO/IEC 42001 governance. Pricing is early and typically per environment or model scope. Several incumbents are acquiring into the space.

Relevant only if you build/deploy AI; aligns with NIST AI RMF / ISO 42001.

Advantages

Secures the AI lifecycle and LLM apps
aligns with NIST AI RMF and ISO/IEC 42001
Lakera/Prompt strong on prompt injection
incumbents acquiring in.

Trade-offs

Very new, churning market with heavy M&A
standards immature
pricing unsettled
only relevant with meaningful AI use.

Vendors & tools in this area

Palo Alto (Protect AI)Commercial

AI/ML security: model scanning, ML supply chain.

HiddenLayerCommercial

ML model threat detection.

Check Point (Lakera)Commercial

LLM / prompt-injection guardrails Acquired by Check Point (closed Q4 2025).

Cisco (Robust Intelligence)Commercial

AI firewall / model validation (Cisco).

SentinelOne (Prompt Security)Commercial

GenAI / LLM application security Acquired by SentinelOne (closed Sep 2025).

GarakOpen source

LLM vulnerability scanner.

NeMo GuardrailsOpen source

NVIDIA LLM guardrails.

RebuffOpen source

Prompt-injection detection — repo archived, no longer maintained.

PromptfooOpen source

Open-source LLM eval and red-teaming CLI, CI/CD-native.

Noma SecurityCommercial

Unified AI/agent discovery, posture management, red teaming, and runtime protection.

GiskardOpen source

Open-source LLM/ML testing library for hallucination, bias, and prompt-injection detection.

Start free trial Browse the full atlas

AI / ML Security (AI-SPM) — Cybersecurity Landscape Atlas — CMDash